Abstract:
|
Often, the timing for an interim or final analysis for a clinical trial will be based on the occurrence of a number of events. For instance, an oncology study's final analysis may occur after a given number of progressions or deaths have happened. In these situations, it can be difficult to determine when these analyses will happen, yet study teams need to know as far in advance as possible to schedule resources, convene data review committees, or prepare public statements. Simplistic projections of these analysis times fail to account for uncertainty in the estimate of the rate of events. We propose a model that fits separate Weibull distributions to model both events (e.g. survival) and early terminations to not only estimate the best time to schedule such an analysis, but also the uncertainty in the scheduling. The model will be illustrated using a realistic scenario and compared to using current event monitoring techniques.
|